Jailbreaking and other Creative Practices in Human-LLM Interaction

Interdisciplinary Workshop at Bielefeld University on August 28, 2025

organized by Ole Pütz, Elena Esposito & Jonas Mieke

We are excited to invite various European researchers from different academic backgrounds to our workshop on creative practices in human-LLM interaction, taking place on August 28 in the Faculty of Sociology (building X) at Bielefeld University! We have put together a great program with exciting presentations and enough time for in-depth discussion of our work-in-progress projects and papers.

This interdisciplinary workshop addresses dynamics, challenges and opportunities of different forms of alignment and misalignment in user interactions with LLM-powered tools.

Jailbreaking practices provide an exemplary starting point, because they reveal both the sophistication of the efforts to contain the misalignment of systems by developers and the creativity of the attempts to bypass them by users — with either beneficial or malicious intentions.

Our discussion solicits contributions that analyse similar and related issues, such as:

  • LLM-powered chatbots in the wild, e.g. customer service chatbots or social bots that pretend to be humans
  • Practices of disagreement, disruption and repair in human-LLM interaction
  • Strategies that users and developers apply to manage conflicts between normative alignment and functionality
  • Conditions of acceptability and understanding in user-LLM interaction across different cases, applications and regulatory regimes
  • Normative assumptions and ideological framings in training or user guidelines by tech companies of different LLMs
  • Mechanisms and practices linking roles, role-play and (mis-)alignment
  • Attributions of trust and responsibility to bots, companies, and users

We welcome both empirical contributions that explore the various aspects of the interaction of users with LLMs-powered tools and theoretical contributions that analyse their communicative and social structures.

Program

We are planning presentations of approximately 20 minutes to allow sufficient time for thorough discussions.

TimeTopicSpeaker
09:00 – 09:30 Welcome and IntroductionOle Pütz, Elena Esposito, Jonas Mieke
09:30 – 10:20 Role-play based Jailbreaking
and the Social Structuring of Artificial Communication
Jonas Mieke, Elena Esposito, Ole Pütz
Bielefeld University
10:20 – 10:30 Coffee break
10:30 – 11:20 Membership Categorisation Analysis, Everyday Sociological Reasoning and Prompt EngineeringPatrik Dahl, William Housley
Cardiff University, Bielefeld University
11:20 – 12:10 Latent culture: Theorizing “values” in human-LLM relationsAnders Kristian Munk
DTU Management (Kongens Lyngby, Copenhagen)
12:10 – 13:10 Lunch break
13:10 – 14:00 Synthetic Reflexivity. The Interweaving of Generative AI and Social AwarenessFederico Pilati
University of Bologna
14:00 – 14:50 Human-LLM Interaction: Empirical Observations and Novel Tools for Collaborative ResearchEirik Lidsheim, Marlou Rasenberg,
Mark Dingemanse, Andreas Liesenfeld
Radboud University Nijmegen
14:50 – 15:10 Coffee break
15:10 – 16:00 Virtual footing: prompting agency in LLM voice interfacesSaul Albert
Loughborough University
16:00 – 17:00 General discussion with all speakers and guestsElena Esposito, Ole Pütz, Jonas Mieke

Important Dates:

  • send papers by: August 10, 2025
  • workshop date: August 28, 2025

Contact

If you have any questions or requests, Jonas Mieke will be happy to help. You can reach him by email at: jonas.mieke@uni-bielefeld.de